Many aspects of human health are affected by the process of genetic adaptation. Somatic cells escape growth control and become malignant. Pathogens evade host defenses and cause disease. Inherited disorders would be more easily associated with mutations by knowing how genetic changes occur. Adaptation can be very rapid, despite very low mutation rates. We propose that experimental genetics has missed a major pathway of genetic adaptation. Genetics focuses on rare large-effect mutations that are easier to identify and simpler to analyze. This ignores the two commonest mutation types (missense and copy number changes). Most point mutations and copy number changes have very small phenotypes and are hard to remove from the population (if deleterious) and hard to detect experimentally (if beneficial). Both types are carried as frequent polymorphisms in populations. When selection is imposed, the resident small effect mutant alleles can amplify under selection and provide an increased phenotype with a greater selective advantage. Amplification provides more copies of the growth-limiting gene and more gene targets for mutational improvement. Once the problem is solved, selection is relaxed, the amplification is lost and only the improved allele is retained Selection thus appears to induce mutagenesis and direct mutations to beneficial sites. We suggest that most gene duplications are initiated by palindromic sequences, which can form snap-back structures and prime repair synthesis leading to a symmetrical tandem inversion duplication (sTID) - abcd-d'c'b'a'-abcd. These duplications amplify under selection and are stabilized by join point deletions that shorten the repeated unit, improve fitness and allow even higher amplification. We will look for the predicted intermediate unstable symmetric TID junctions using a new non-selective trapping methods. We suggest that the initiating palindromic sequences are activated by transcription, which allows secondary (snap back) structures to form in the non- template DNA strand of R-loops. This is most likely to occur during transcription of A/T-rich, G/C skewed sequences and in operons that include palindromic dispersed repeats - REP(BIMES), ERIC, AelRep and LasRep. Other potential sources of TID-initiating palindromes are the highly- transcribed, palindromic ribosomal RNA (rrn) genes.